Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A device for verifying image-related information of a medical image, said device comprising: an image input for obtaining a medical image and associated image-related information accompanying or arising out of the medical image, a translator for translating one or more pieces of the image-related information into expected anatomical feature information, a detector for searching the expected anatomical feature information in the obtained medical image or for detecting one or more given anatomical features in the obtained medical image and comparing them with the expected anatomical feature information, an evaluator for evaluating the result of the detector to generate a consistency measure indicating the consistency of the image-related information with the associated medical image, and an output interface for outputting an inconsistency indication if the generated consistency measure is lower than a predetermined consistency threshold.
2. The device as claimed in claim 1 , wherein said detector is configured to generate a confidence measure indicating the confidence, with which one or more pieces of expected anatomical feature information have been detected in the obtained medical image or with which one or more given anatomical features in the obtained medical image correspond with the expected anatomical feature information, and wherein said evaluator is configured to use the generated confidence measure in the generation of the consistency measure.
3. The device as claimed in claim 2 , wherein said detector is configured to perform the detection separately for expected anatomical feature information from two or more pieces of image-related information to generate a confidence measure per piece of image-related information and wherein said evaluator is configured to combine the generated confidence measures for the generation of the consistency measure.
4. The device as claimed in claim 2 , wherein said translator is configured to separately translate each of two or more pieces of image-related information into separate pieces of expected anatomical feature information, wherein said detector is configured to perform the detection separately for two or more pieces of expected anatomical feature information to generate a confidence measure per piece of image-related information and wherein said evaluator is configured to combine the generated confidence measures for the generation of the consistency measure.
This invention relates to medical imaging systems that analyze anatomical features in images to detect inconsistencies. The problem addressed is the need for accurate and reliable detection of anatomical features in medical images, particularly when multiple pieces of image-related information are involved. The invention improves upon prior systems by enhancing the detection process through a multi-step translation and evaluation mechanism. The device includes a translator that processes two or more pieces of image-related information, such as different imaging modalities or regions of interest, and converts each into separate pieces of expected anatomical feature information. A detector then independently analyzes each piece of expected anatomical feature information to generate a confidence measure for each. These confidence measures reflect the likelihood that the detected features match the expected anatomical structures. An evaluator then combines these individual confidence measures to produce a consistency measure, which indicates how well the detected features align across the different pieces of image-related information. This approach improves detection accuracy by leveraging multiple sources of information and ensuring consistency across them. The system is particularly useful in medical diagnostics where precise identification of anatomical features is critical.
5. The device as claimed in claim 1 , wherein said translator is configured to translate one or more pieces of image-related information into expected anatomical feature information including the number and/or kind of anatomical features, expected image properties, distribution and/or location of one or more expected anatomical features, and/or variability in the spatial relations of one or more expected anatomical features.
This invention relates to medical imaging systems that analyze anatomical features in images. The problem addressed is the need for accurate and automated identification of anatomical structures in medical images, which is challenging due to variability in patient anatomy, imaging conditions, and feature distribution. The invention provides a device with a translator component that processes image-related information to generate expected anatomical feature data. This includes determining the number and type of anatomical features present, their expected image properties, spatial distribution, and variability in their relative positions. The translator uses this information to enhance image analysis, improving the accuracy of feature detection and reducing errors caused by anatomical variations. The system may also incorporate prior knowledge or learned models to refine predictions. By translating raw image data into structured anatomical expectations, the device supports applications such as automated diagnosis, surgical planning, and image-guided interventions. The invention aims to standardize and automate the interpretation of medical images, reducing reliance on manual analysis and improving consistency in clinical workflows.
6. The device as claimed in claim 1 , further comprising a controller for controlling the detector and the evaluator to repeat their operations one or more times with different pieces of expected anatomical feature information if the generated consistency is lower than the predetermined consistency threshold or another predetermined consistency threshold.
7. The device as claimed in claim 1 , wherein said translator is configured to access a database storing separate expected anatomical feature information for a plurality of pieces of image-related information for translating one or more pieces of image-related information into expected anatomical feature information.
8. The device as claimed in claim 7 , wherein said translator is configured to access a database storing the number and/or kind of anatomical features, expected image properties, distribution and/or location of one or more expected anatomical features, and/or variability in the spatial relations of one or more expected anatomical features.
9. The device as claimed in claim 7 , wherein said translator is configured to access a database storing separate expected anatomical feature information obtained by learning from training medical images and correctly associated image-related information.
10. The device as claimed in claim 6 , wherein said detector is configured to search alternative anatomical feature information in the obtained medical image or to detect one or more given anatomical features in the obtained medical image and compare them with alternative anatomical feature information, wherein said evaluator is configured to evaluate the result of the detector to generate an alternative consistency measure indicating the consistency of the alternative image-related information with the associated medical image, and wherein said output interface is configured to output an inconsistency indication if the previously generated consistency measure and the alternative consistency measure are lower than a predetermined consistency threshold.
11. The device as claimed in claim 1 , wherein said image input is configured to obtain a medical image file comprising a data part including the medical image and a header including associated image-related information, in particular a medical image file in accordance with the DICOM standard.
This invention relates to medical imaging systems designed to process and analyze medical image files. The primary problem addressed is the efficient handling of medical image data, particularly in formats like DICOM, which include both image data and metadata in a structured header. The invention describes a device that processes these files by extracting and utilizing the header information alongside the image data. The device includes an image input module configured to obtain medical image files containing a data part with the actual medical image and a header part with associated metadata, such as patient information, imaging parameters, and other image-related details. The system is specifically adapted to work with DICOM-standard files, which are widely used in medical imaging. The device may further include processing components to analyze or manipulate the image data based on the metadata, ensuring accurate and context-aware medical imaging workflows. This approach improves data management, interoperability, and diagnostic accuracy by leveraging standardized file structures and metadata. The invention aims to enhance medical imaging systems by ensuring seamless integration of image data and metadata, facilitating better clinical decision-making and workflow efficiency.
12. The device as claimed in claim 1 , wherein said output interface is configured to output an inconsistency indication in the form of an alert, in particular including the degree of inconsistency.
13. A method for verifying image-related information of a medical image, said method comprising: obtaining a medical image and associated image-related information arising out of the medical image, translating one or more pieces of the image-related information into expected anatomical feature information, searching the expected anatomical feature information in the obtained medical image or detecting one or more given anatomical features in the obtained medical image, which are then compared with the expected anatomical feature information, evaluating the result of the step of searching or detecting to generate a consistency measure indicating the consistency of the image-related information with the associated medical image, and outputting an inconsistency indication if the generated consistency measure is lower than a predetermined consistency threshold.
14. A computer program comprising program code means for causing a computer to carry out the steps of the method as claimed in claim 13 when said computer program is carried out on the computer.
15. The method as claimed in claim 13 , wherein the associated image-related information is a body part, image laterality, imaging position of the subject, view position, type of imaging system, or parameter or setting of image acquisition equipment.
16. The method as claimed in claim 13 wherein the expected anatomical feature information is a landmark, contour, distance information, size information, anatomical feature identification, shape descriptor, or model pose.
17. The method as claimed in claim 13 , wherein the consistency measure indicates a correctness of the image-related information.
This invention relates to evaluating the consistency of image-related information in a machine learning system, particularly for assessing the accuracy of data used in training or inference tasks. The method involves analyzing image-related information, such as labels, annotations, or metadata, to determine its consistency with the visual content of the image. A consistency measure is computed to quantify how well the provided information aligns with the image, where a higher measure indicates greater correctness. The method may involve comparing extracted features from the image against the provided information or using a trained model to assess alignment. The consistency measure can be used to filter or prioritize data, improve model training, or detect errors in annotations. The approach helps ensure high-quality datasets by identifying discrepancies between image content and associated metadata, reducing errors in machine learning applications. The system may also include preprocessing steps to standardize the image-related information before consistency evaluation. The method is applicable in various domains, including computer vision, autonomous systems, and data validation pipelines.
18. A method for verifying the correctness of image-related information of a medical image, said method comprising: receiving a medical image and associated image-related information derived from the medical image, translating one or more pieces of the image-related information into expected anatomical feature information, searching the expected anatomical feature information in the received medical image or detecting one or more given anatomical features in the received medical image, which are then compared with the expected anatomical feature information, evaluating the result of the step of searching or detecting to generate a consistency measure indicating the consistency of the image-related information with the associated medical image, and storing the image-related information if the consistency measure is higher than a predetermined consistency threshold and verifies the correctness of the image-related information.
19. The method according to claim 18 , wherein the associated image-related information is received as a part of or during the acquisition of the medical image.
20. The method according to claim 13 , wherein the associated image-related information is obtained as a part of or during the acquisition of the medical image.
Unknown
March 2, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.